1 /*
2 * Copyright (c) 2015, 2019, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #ifndef SHARE_GC_SHARED_TASKQUEUE_INLINE_HPP
26 #define SHARE_GC_SHARED_TASKQUEUE_INLINE_HPP
27
28 #include "gc/shared/taskqueue.hpp"
29 #include "memory/allocation.inline.hpp"
30 #include "oops/oop.inline.hpp"
31 #include "runtime/atomic.hpp"
32 #include "runtime/orderAccess.hpp"
33 #include "utilities/debug.hpp"
34 #include "utilities/stack.inline.hpp"
35
36 template <class T, MEMFLAGS F>
37 inline GenericTaskQueueSet<T, F>::GenericTaskQueueSet(uint n) : _n(n) {
38 typedef T* GenericTaskQueuePtr;
39 _queues = NEW_C_HEAP_ARRAY(GenericTaskQueuePtr, n, F);
40 for (uint i = 0; i < n; i++) {
41 _queues[i] = NULL;
42 }
43 }
44
45 template <class T, MEMFLAGS F>
46 inline GenericTaskQueueSet<T, F>::~GenericTaskQueueSet() {
47 FREE_C_HEAP_ARRAY(T*, _queues);
48 }
49
50 template<class E, MEMFLAGS F, unsigned int N>
51 inline void GenericTaskQueue<E, F, N>::initialize() {
52 _elems = ArrayAllocator<E>::allocate(N, F);
53 }
54
55 template<class E, MEMFLAGS F, unsigned int N>
56 inline GenericTaskQueue<E, F, N>::~GenericTaskQueue() {
57 ArrayAllocator<E>::free(const_cast<E*>(_elems), N);
58 }
59
60 template<class E, MEMFLAGS F, unsigned int N>
61 bool GenericTaskQueue<E, F, N>::push_slow(E t, uint dirty_n_elems) {
62 if (dirty_n_elems == N - 1) {
63 // Actually means 0, so do the push.
64 uint localBot = _bottom;
65 // g++ complains if the volatile result of the assignment is
66 // unused, so we cast the volatile away. We cannot cast directly
67 // to void, because gcc treats that as not using the result of the
68 // assignment. However, casting to E& means that we trigger an
69 // unused-value warning. So, we cast the E& to void.
70 (void)const_cast<E&>(_elems[localBot] = t);
71 OrderAccess::release_store(&_bottom, increment_index(localBot));
72 TASKQUEUE_STATS_ONLY(stats.record_push());
73 return true;
74 }
75 return false;
76 }
77
78 template<class E, MEMFLAGS F, unsigned int N> inline bool
79 GenericTaskQueue<E, F, N>::push(E t) {
80 uint localBot = _bottom;
81 assert(localBot < N, "_bottom out of range.");
82 idx_t top = _age.top();
83 uint dirty_n_elems = dirty_size(localBot, top);
84 assert(dirty_n_elems < N, "n_elems out of range.");
85 if (dirty_n_elems < max_elems()) {
86 // g++ complains if the volatile result of the assignment is
87 // unused, so we cast the volatile away. We cannot cast directly
88 // to void, because gcc treats that as not using the result of the
89 // assignment. However, casting to E& means that we trigger an
90 // unused-value warning. So, we cast the E& to void.
91 (void) const_cast<E&>(_elems[localBot] = t);
92 OrderAccess::release_store(&_bottom, increment_index(localBot));
93 TASKQUEUE_STATS_ONLY(stats.record_push());
94 return true;
95 } else {
96 return push_slow(t, dirty_n_elems);
97 }
98 }
99
100 template <class E, MEMFLAGS F, unsigned int N>
101 inline bool OverflowTaskQueue<E, F, N>::push(E t)
102 {
103 if (!taskqueue_t::push(t)) {
104 overflow_stack()->push(t);
105 TASKQUEUE_STATS_ONLY(stats.record_overflow(overflow_stack()->size()));
106 }
107 return true;
108 }
109
110 template <class E, MEMFLAGS F, unsigned int N>
111 inline bool OverflowTaskQueue<E, F, N>::try_push_to_taskqueue(E t) {
112 return taskqueue_t::push(t);
113 }
114
115 // pop_local_slow() is done by the owning thread and is trying to
116 // get the last task in the queue. It will compete with pop_global()
117 // that will be used by other threads. The tag age is incremented
118 // whenever the queue goes empty which it will do here if this thread
119 // gets the last task or in pop_global() if the queue wraps (top == 0
120 // and pop_global() succeeds, see pop_global()).
121 template<class E, MEMFLAGS F, unsigned int N>
122 bool GenericTaskQueue<E, F, N>::pop_local_slow(uint localBot, Age oldAge) {
123 // This queue was observed to contain exactly one element; either this
124 // thread will claim it, or a competing "pop_global". In either case,
125 // the queue will be logically empty afterwards. Create a new Age value
126 // that represents the empty queue for the given value of "_bottom". (We
127 // must also increment "tag" because of the case where "bottom == 1",
128 // "top == 0". A pop_global could read the queue element in that case,
129 // then have the owner thread do a pop followed by another push. Without
130 // the incrementing of "tag", the pop_global's CAS could succeed,
131 // allowing it to believe it has claimed the stale element.)
132 Age newAge((idx_t)localBot, oldAge.tag() + 1);
133 // Perhaps a competing pop_global has already incremented "top", in which
134 // case it wins the element.
135 if (localBot == oldAge.top()) {
136 // No competing pop_global has yet incremented "top"; we'll try to
137 // install new_age, thus claiming the element.
138 Age tempAge = _age.cmpxchg(newAge, oldAge);
139 if (tempAge == oldAge) {
140 // We win.
141 assert(dirty_size(localBot, _age.top()) != N - 1, "sanity");
142 TASKQUEUE_STATS_ONLY(stats.record_pop_slow());
143 return true;
144 }
145 }
146 // We lose; a completing pop_global gets the element. But the queue is empty
147 // and top is greater than bottom. Fix this representation of the empty queue
148 // to become the canonical one.
149 _age.set(newAge);
150 assert(dirty_size(localBot, _age.top()) != N - 1, "sanity");
151 return false;
152 }
153
154 template<class E, MEMFLAGS F, unsigned int N> inline bool
155 GenericTaskQueue<E, F, N>::pop_local(volatile E& t, uint threshold) {
156 uint localBot = _bottom;
157 // This value cannot be N-1. That can only occur as a result of
158 // the assignment to bottom in this method. If it does, this method
159 // resets the size to 0 before the next call (which is sequential,
160 // since this is pop_local.)
161 uint dirty_n_elems = dirty_size(localBot, _age.top());
162 assert(dirty_n_elems != N - 1, "Shouldn't be possible...");
163 if (dirty_n_elems <= threshold) return false;
164 localBot = decrement_index(localBot);
165 _bottom = localBot;
166 // This is necessary to prevent any read below from being reordered
167 // before the store just above.
168 OrderAccess::fence();
169 // g++ complains if the volatile result of the assignment is
170 // unused, so we cast the volatile away. We cannot cast directly
171 // to void, because gcc treats that as not using the result of the
172 // assignment. However, casting to E& means that we trigger an
173 // unused-value warning. So, we cast the E& to void.
174 (void) const_cast<E&>(t = _elems[localBot]);
175 // This is a second read of "age"; the "size()" above is the first.
176 // If there's still at least one element in the queue, based on the
177 // "_bottom" and "age" we've read, then there can be no interference with
178 // a "pop_global" operation, and we're done.
179 idx_t tp = _age.top(); // XXX
180 if (size(localBot, tp) > 0) {
181 assert(dirty_size(localBot, tp) != N - 1, "sanity");
182 TASKQUEUE_STATS_ONLY(stats.record_pop());
183 return true;
184 } else {
185 // Otherwise, the queue contained exactly one element; we take the slow
186 // path.
187
188 // The barrier is required to prevent reordering the two reads of _age:
189 // one is the _age.get() below, and the other is _age.top() above the if-stmt.
190 // The algorithm may fail if _age.get() reads an older value than _age.top().
191 OrderAccess::loadload();
192 return pop_local_slow(localBot, _age.get());
193 }
194 }
195
196 template <class E, MEMFLAGS F, unsigned int N>
197 bool OverflowTaskQueue<E, F, N>::pop_overflow(E& t)
198 {
199 if (overflow_empty()) return false;
200 t = overflow_stack()->pop();
201 return true;
202 }
203
204 template<class E, MEMFLAGS F, unsigned int N>
205 bool GenericTaskQueue<E, F, N>::pop_global(volatile E& t) {
206 Age oldAge = _age.get();
207 // Architectures with weak memory model require a barrier here
208 // to guarantee that bottom is not older than age,
209 // which is crucial for the correctness of the algorithm.
210 #ifndef CPU_MULTI_COPY_ATOMIC
211 OrderAccess::fence();
212 #endif
213 uint localBot = OrderAccess::load_acquire(&_bottom);
214 uint n_elems = size(localBot, oldAge.top());
215 if (n_elems == 0) {
216 return false;
217 }
218
219 // g++ complains if the volatile result of the assignment is
220 // unused, so we cast the volatile away. We cannot cast directly
221 // to void, because gcc treats that as not using the result of the
222 // assignment. However, casting to E& means that we trigger an
223 // unused-value warning. So, we cast the E& to void.
224 (void) const_cast<E&>(t = _elems[oldAge.top()]);
225 Age newAge(oldAge);
226 newAge.increment();
227 Age resAge = _age.cmpxchg(newAge, oldAge);
228
229 // Note that using "_bottom" here might fail, since a pop_local might
230 // have decremented it.
231 assert(dirty_size(localBot, newAge.top()) != N - 1, "sanity");
232 return resAge == oldAge;
233 }
234
235 inline int randomParkAndMiller(int *seed0) {
236 const int a = 16807;
237 const int m = 2147483647;
238 const int q = 127773; /* m div a */
239 const int r = 2836; /* m mod a */
240 STATIC_ASSERT(sizeof(int) == 4);
241 int seed = *seed0;
242 int hi = seed / q;
243 int lo = seed % q;
244 int test = a * lo - r * hi;
245 if (test > 0) {
246 seed = test;
247 } else {
248 seed = test + m;
249 }
250 *seed0 = seed;
251 return seed;
252 }
253
254 template<class E, MEMFLAGS F, unsigned int N>
255 int GenericTaskQueue<E, F, N>::next_random_queue_id() {
256 return randomParkAndMiller(&_seed);
257 }
258
259 template<class T, MEMFLAGS F> bool
260 GenericTaskQueueSet<T, F>::steal_best_of_2(uint queue_num, E& t) {
261 if (_n > 2) {
262 T* const local_queue = _queues[queue_num];
263 uint k1 = queue_num;
264
265 if (local_queue->is_last_stolen_queue_id_valid()) {
266 k1 = local_queue->last_stolen_queue_id();
267 assert(k1 != queue_num, "Should not be the same");
268 } else {
269 while (k1 == queue_num) {
270 k1 = local_queue->next_random_queue_id() % _n;
271 }
272 }
273
274 uint k2 = queue_num;
275 while (k2 == queue_num || k2 == k1) {
276 k2 = local_queue->next_random_queue_id() % _n;
277 }
278 // Sample both and try the larger.
279 uint sz1 = _queues[k1]->size();
280 uint sz2 = _queues[k2]->size();
281
282 uint sel_k = 0;
283 bool suc = false;
284
285 if (sz2 > sz1) {
286 sel_k = k2;
287 suc = _queues[k2]->pop_global(t);
288 } else if (sz1 > 0) {
289 sel_k = k1;
290 suc = _queues[k1]->pop_global(t);
291 }
292
293 if (suc) {
294 local_queue->set_last_stolen_queue_id(sel_k);
295 } else {
296 local_queue->invalidate_last_stolen_queue_id();
297 }
298
299 return suc;
300 } else if (_n == 2) {
301 // Just try the other one.
302 uint k = (queue_num + 1) % 2;
303 return _queues[k]->pop_global(t);
304 } else {
305 assert(_n == 1, "can't be zero.");
306 return false;
307 }
308 }
309
310 template<class T, MEMFLAGS F> bool
311 GenericTaskQueueSet<T, F>::steal(uint queue_num, E& t) {
312 for (uint i = 0; i < 2 * _n; i++) {
313 TASKQUEUE_STATS_ONLY(queue(queue_num)->stats.record_steal_attempt());
314 if (steal_best_of_2(queue_num, t)) {
315 TASKQUEUE_STATS_ONLY(queue(queue_num)->stats.record_steal());
316 return true;
317 }
318 }
319 return false;
320 }
321
322 template <unsigned int N, MEMFLAGS F>
323 inline typename TaskQueueSuper<N, F>::Age TaskQueueSuper<N, F>::Age::cmpxchg(const Age new_age, const Age old_age) volatile {
324 return Atomic::cmpxchg(new_age._data, &_data, old_age._data);
325 }
326
327 template<class E, MEMFLAGS F, unsigned int N>
328 template<class Fn>
329 inline void GenericTaskQueue<E, F, N>::iterate(Fn fn) {
330 uint iters = size();
331 uint index = _bottom;
332 for (uint i = 0; i < iters; ++i) {
333 index = decrement_index(index);
334 fn(const_cast<E&>(_elems[index])); // cast away volatility
335 }
336 }
337
338
339 #endif // SHARE_GC_SHARED_TASKQUEUE_INLINE_HPP